Research

Research — Hybrid Router + Fallback

The runtime orchestrator for the research domain. Architecture C: deterministic classification → specialist delegation OR own plan-decompose-search-synthesize-cite workflow.

Portability

Requires WebSearch + WebFetch for the fallback workflow; specialist skills (pulse, grants, litreview, syllabus, patent, dossier) must be present for delegation to work. Node.js with docx package required if Q2 = document mode. Works in Claude Code CLI natively. In Claude.ai with web tools + Code Execution, the workflow is supported.

Distinct From engineering/autoresearch-agent

These two skills share the word "research" but serve completely different use cases:

• research/research/ (this skill) — research-query router + fallback workflow ("Research X")
• engineering/autoresearch-agent/ — Karpathy's autonomous file-optimization experiment loop ("Make this code faster")

No overlap. They coexist.

Hybrid Architecture (C)

Every invocation produces one of three outcomes:

1. Delegation — Classified as specialist-domain. Routes there. User sees the specialist's output.
2. Fallback execution — Classified as general research. Runs own plan → search → synthesize workflow.
3. Clarification request — Classification ambiguous. Asks one forcing question to disambiguate, then routes.

The skill never silently runs its fallback when a specialist would have done better. Routing transparency is what makes the hybrid architecture trustworthy.

Specialist Registry

Agent Integrity Rules

This skill obeys the research-pack convention:

• Execution discipline (fallback only): Sequential searches. 1 q/sec rate limit. Confirm response received before next call.
• Source discipline: Cite only sources returned by this session's tool calls. Training knowledge labeled [Background — not from search] and excluded from counts.
• Three-count tracking (fallback only): Queries sent / sources received / sources cited.
• Retry policy: On failure → wait 3s → retry once → log. After 3 consecutive failures: stop, alert user.
• Plan-tier detection: If delegated to Consensus-using specialist, that specialist handles detection. In fallback mode, surface any rate-limit signals.
• Routing discipline: Never delegate silently. Always state the decision + accept override.

Phase 1: Grill-Me Intake (2–4 Questions)

Intake is intentionally minimal — the goal is to route fast, not to interrogate. One question per turn.

Q1 (always) — Research question

What's the research question? State it in 1–2 sentences. Specific is better than broad — "AI for healthcare" gets you a vague survey; "How are health systems integrating LLM-based clinical decision support in 2026?" gets you a useful answer.

Why I'm asking: Specificity dictates classification accuracy and search precision. A vague question routes to fallback; a specific question often matches a specialist cleanly.

Refuse mush. If user says "research AI", push back once: "What about AI specifically — adoption, safety, capability, funding, regulation, comparison? Pick an angle."

Q2 (always) — Output preference

What output do you want? Pick one:

1. Quick chat briefing (5-min read, markdown in chat)
2. Standalone document (.docx with citations, shareable)

Why I'm asking: Document mode triggers deeper search budgets and full audit logs. Chat mode optimizes for fast delivery.

Forcing choice.

Q3 (asked only if classification ambiguous — ≤1 signal) — Domain disambiguation

Quick clarification — pick the closest match:

1. Academic literature (papers, peer-reviewed)
2. Industry / trends (what's the buzz, news, sentiment)
3. Specific entity (a company, person, organization)
4. Technology / patents (prior art, IP landscape)
5. Grant funding (NIH, foundations)
6. Course material (syllabus or curriculum)
7. None of the above — run general research

Why I'm asking: I couldn't classify confidently from your question alone. This routes you to the right specialist or confirms general-research fallback.

Skip if Q1 + Q2 produced clear specialist match (≥2 signals).

Q4 (asked only if Q3 was needed AND user picked "none of the above") — General-research scope

For general research, what's your time horizon — quick scan (5 searches) or thorough (15 searches)?

Why I'm asking: General research has no specialist budget; you pick it. Quick is good for "what's the lay of the land". Thorough is for "I'll make a decision based on this".

Skip if a specialist took over.

Stop condition: After Q4 (or earlier if dependency skips applied), commit and start Phase 2. Most invocations exit intake after Q1 + Q2.

Phase 2: Deterministic Classification

This is deterministic, not LLM-reasoned — for speed, debuggability, and consistency.

SIGNALS = {
  pulse:    ["reddit", "hn", "hacker news", "x.com", "twitter", "buzz",
             "sentiment", "trending", "what are people saying",
             "what's happening", "the conversation around",
             "pulse on", "take the pulse", "current conversation"],
  grants:   ["nih", "grant", "grants for", "r01", "r21", "k-award", "reporter",
             "nosi", "funding", "fda", "study section", "principal investigator"],
  litreview:["literature review", "lit review", "litreview", "pico", "spider",
             "systematic review", "review papers on", "research papers on",
             "papers about", "meta-analysis"],
  syllabus: ["syllabus", "course outline", "curriculum", "reading list",
             "for my class", "for my students", "course material"],
  patent:   ["prior art", "fto", "freedom to operate", "patent",
             "patent landscape", "invention", "novelty search",
             "patent search", "ip landscape"],
  dossier:  ["dossier on", "due diligence", "background check",
             "prep me for", "competitor research", "investor diligence",
             "interview prep", "research my competitor", "background on"]
}

# Signals are case-insensitive literal phrases (multi-word substring match).
# Bracketed placeholders (e.g., "research [company]") are intentionally NOT
# signals — they over-trigger on generic "research X" queries that should
# fall back to general research, not auto-route to dossier. Specific phrases
# pair the verb with the noun ("dossier on", "background on") and route reliably.

For each specialist S:
  score[S] = count of SIGNALS[S] phrases matched in question (case-insensitive substring)

if max(score) >= 2:
  route_to = argmax(score)                                  # high confidence
elif max(score) == 1 and only one specialist has score 1:
  route_to = that specialist                                # weak match, single specialist
else:
  route_to = "fallback"                                     # ambiguous or no match — ask Q3

Implementation: scripts/classifier.py --question "..." returns the routing decision + matched signals + per-specialist scores. Use it; don't re-implement.

Phase 3a: Specialist Delegation (≥2 signals OR single weak match)

When delegating:

1. Pass the user's question verbatim plus the output preference (Q2)
2. Let the specialist run its own grill-me intake — do NOT pre-answer specialist questions
3. Return specialist output as the user-visible result
4. Tag the result with [Delegated to: research → {specialist}] in the chat output so the user knows what skill produced it
5. Tag the audit log via scripts/routing_transparency_logger.py --action record_delegation

Phase 3b: Own Fallback Workflow

If routing produced no specialist match, run the 8-step fallback.

Step 1: Decompose

Break the research question into 3–5 sub-questions. Use the framework: what / why / how / who / what's next. Show the decomposition to the user before searching. Use scripts/fallback_decomposer.py --question "..." for a deterministic starting point.

Step 2: Source Selection

For each sub-question, choose source(s) deterministically:

• Recency-sensitive → WebSearch + WebFetch + (optionally Reddit/HN if signal)
• Technical specs / docs → WebSearch + WebFetch
• Academic → Consensus MCP if connected; otherwise WebSearch with scholar.google.com site filter
• Data / numbers → WebSearch for sources; then WebFetch for primary documents
• Person / company entity-level → consider routing to dossier (offer override)

Step 3: Search

Sequential per sub-question. 1 q/sec etiquette. Per source: 2–4 queries, broad-to-narrow.

Step 4: Read + Extract

For each result that looks high-signal: WebFetch and extract the relevant section. Note the source URL.

Step 5: Synthesize

Per sub-question: 2–4 paragraphs answering it with inline citations. Surface disagreement when sources disagree.

Step 6: Cross-Cutting Patterns

After per-sub-question synthesis: 1–2 paragraphs of patterns across sub-questions — consensus, controversy, gaps.

Step 7: Output

Markdown brief by default (Q2 choice). DOCX if user picked document mode.

Step 8: Audit Log

Three-count summary (sent / received / cited) + per-source list with reliability tier (primary / secondary / tertiary).

Routing Transparency Protocol (Mandatory)

After classification, the skill always:

1. States the decision in one sentence: "Routing to litreview because you mentioned PICO and meta-analysis (2 signals)."
2. Offers override: "If you want general research instead OR a different specialist, say so now. Otherwise proceeding in 5 seconds."
3. Waits 1 turn for confirmation (or auto-proceeds after 5s in interactive contexts).
4. If user overrides → accept, re-route, log the override via routing_transparency_logger.py --action record_override.

Never delegates silently. This is the trust-building property that makes the hybrid pattern work.

Output Format

Markdown brief (Q2 = quick chat briefing)

# [Research Question] — Briefing
*Generated: [DATE] | Routed: [delegated specialist | fallback]*

## TL;DR
[2-3 sentences]

## Findings
### [Sub-question 1]
[2-4 paragraphs with inline citations]

### [Sub-question 2]
...

## Cross-Cutting Patterns
[1-2 paragraphs]

## Sources
[Numbered list with hyperlinks, reliability tier per source]

## Audit
[Three counts + per-source tier + failures]

DOCX (Q2 = standalone document)

Use the standard research-pack DOCX patterns: Arial 12pt, navy headings, blue table headers, hyperlinked sources, mandatory audit log section. Reference the docx skill for setup.

Audit Log Requirement (Fallback Mode)

Queries sent:        N
Sources received:    M
Sources cited:       K
Failures:            F (3-consecutive-failures triggered: yes/no)
Per-source tier:     [URL — primary | secondary | tertiary]
Routing decision:    fallback (no specialist matched)
Sub-questions:       [list]

All routing decisions + overrides also logged to ~/.research_sessions/<session>.json via routing_transparency_logger.py.

Failure Modes

Anti-Patterns Rejected

• LLM-reasoned classification (must be deterministic keyword + intent matching)
• Silent delegation (always surface routing decision)
• Refusing to route to a specialist when ≥2 signals match
• Routing to a specialist when classification is genuinely ambiguous (≤1 signal across all)
• Pre-answering the specialist's grill-me intake (let it run its own)
• Running fallback when a specialist would clearly do better
• Fabricating sources in fallback when search is thin
• Skipping audit log in fallback mode
• Treating "dossier on [company]" as fallback when dossier is the right specialist (the verb-noun-paired phrase, not the generic "research X" form, is what routes)
• Treating "what are people saying about X" as fallback when pulse is the right specialist
• Auto-routing generic "research [topic]" queries to a specialist when the user hasn't paired the verb with a specialist-specific noun (e.g., "research Microsoft" alone is ambiguous — could be dossier or general; ask Q3 instead of guessing)

Tooling

Python (stdlib only)

• scripts/classifier.py — Deterministic SIGNALS matching → routing decision + per-specialist score + matched phrases. --question "..." --output json.
• scripts/routing_transparency_logger.py — JSON-backed audit log at ~/.research_sessions/<session>.json. Records every routing decision, override, and delegation handoff.
• scripts/fallback_decomposer.py — Heuristic question → 3–5 sub-questions using what / why / how / who / what's next framework.

Reference Docs (each cites 7+ authoritative sources)

• references/hybrid_router_architecture.md — router-vs-run trade-offs + routing transparency principle
• references/deterministic_classification_canon.md — why keyword > LLM-reasoned for routing
• references/fallback_workflow_canon.md — plan-decompose-search-synthesize methodology

Dependencies

• WebSearch + WebFetch — Required for fallback workflow
• Specialist skills — Required for delegation: pulse, grants, litreview, syllabus, patent, dossier. If a specialist is missing, the router skips it in classification and routes to fallback instead.
• Node.js docx library — Required if user picks document output (Q2 = standalone)
• Consensus MCP — Optional; used in fallback if academic sub-questions surface

Trigger Phrases

• "research [topic]"
• "look into [topic]"
• "what do we know about [topic]"
• "investigate [topic]"
• "find me information on [topic]"
• "do some research on [topic]"
• "I need to understand [topic]"
• Any research request that doesn't obviously match a more-specific specialist

Version: 1.0.0 Source spec: megaprompts/13-research-megaprompt.md Build pattern: Path B (direct conversion)